Part 3 · 2 chapters · ~12 min

Cognitive Biases That Hit Engineers

Confirmation bias in debugging, anchoring in estimates, the planning fallacy and reference-class forecasting, sunk cost in projects, survivorship and availability biases, overconfidence, and team practices that counter them structurally.

7

Six biases and their counters

Knowing about biases does little on its own. What helps is process: practices that make the biased path harder (silent estimation, written hypotheses, premortems, reference classes).

BIASES THAT HIT ENGINEERS
and the practice that counters each
confirmation biasSeeing evidence for the cause youalready suspect. Counter: listwhat would disprove it.anchoringThe first estimate sticks.Counter: estimate independentlybefore discussing.planning fallacyUnderestimating time for your ownplans. Counter: reference-classforecasting.sunk costContinuing because of what wasspent. Counter: decide on futurecosts and benefits only.survivorship biasLearning only from survivors.Counter: ask what happened to theones that failed.availability biasOverweighting vivid recent events.Counter: look at the base ratedata.
swipe the figure sideways, or tap expand for full screen
1/6
confirmation
During incidents, the first hypothesis attracts all the evidence. Write down what you would expect to see if you were wrong, and look for it.
seek disconfirming evidenceincidents are where it bites
8

Overconfidence and calibration

Engineers are often overconfident about estimates and root causes. Calibration is a trainable skill: when you say you are 90% sure, you should be right about 9 times in 10. Practise by attaching probabilities to predictions ("70% this ships by the 30th") and checking them later; a decision journal (part 6) makes this routine.